myxce/src/lib/kvmgr.c
Rongsong Shen dcf9f39c22 fix(job-array-single-wal-record): close bug-1443's bulk-insert partial-visibility gap
restore_job_array()'s bulk-insert pass could fail partway through
(key_add() on realloc growth exhaustion), leaving an already-inserted
prefix permanently visible in c->jobmgr but excluded from
out_new_jobs -- so those elements existed and were individually
addressable but never reached HOOK_DETAIL_JOB_NEW, never scheduled or
pushed, not on this attempt or any later successful retry. A literal
violation of Requirements 2.3 and 9.4, tracked as bug-1443.

Adds kvmgr_reserve()/xzarray_reserve(), new purely-additive primitives
in shared kvmgr.c/zarray.c (following the existing kvmgr_capture_keys()
template -- bare extern function, no RTTI vtable slot, caller-holds-
write-lock contract), that pre-grow c->jobmgr's backing store before
the bulk-insert pass begins, closing the dominant N-scaling failure
trigger structurally.

Found a genuine off-by-one in the existing, unmodified
xzarray_ensure_items(): its strict less-than sufficiency check means
reserving exactly array_size and then performing array_size real
stores still triggers one more growth cycle on the last store.
Reserving array_size + 1 is the exact minimal correct compensation,
independently re-derived and confirmed by review with concrete
arithmetic.

For the now-much-narrower residual case (kvtab_add()'s own small
per-entry allocation, or a content collision -- both non-N-scaling,
honestly documented as still theoretically possible): restore_job_array()
now populates out_new_jobs with whatever prefix was genuinely inserted
before a failure, and apply_committed_entry() fires HOOK_DETAIL_JOB_NEW
for that non-empty partial prefix even when the overall return is
negative, while still propagating the negative rc so the LSN correctly
halts and the Live Apply Resource Fence still engages for the
remainder. A genuinely-inserted element is never left un-pushed again.

Task 7.1 of job-array-single-wal-record (Requirements 2.3, 9.4, 9.6).
2026-07-16 15:08:02 +00:00

436 lines
11 KiB
C

#include <errno.h>
#include <inttypes.h>
#include <rtti.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "htab.h"
#include "kv.h"
#include "kvmgr.h"
#include "mylock.h"
#include "persist.h"
#include "utils.h"
#include "xce/types.h"
#include "zarray.h"
#include "controller.capnp.h"
typedef struct {
my_rwlock_t lock;
} xkvmgr_private_t;
typedef struct {
kvmgr_key_type_t key_type;
uint32_t nslots;
var kvtab;
var data;
} xkvmgr_t;
static size_t xkvmgr_datasize(va_list ap)
{
size_t sz = sizeof(xkvmgr_t);
return sz;
}
static int xkvmgr_init(meta_obj_t* obj, va_list ap)
{
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
kvmgr_params_t* params = va_arg(ap, kvmgr_params_t*);
kv_params_t kv_params;
xzarray_params_t data_params;
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
if ((params == NULL)
|| ((params->key_type != KVMGR_KEY_INT)
&& (params->key_type != KVMGR_KEY_STRING))
|| (params->nslots <= 1) || (params->delta < 1)) {
return -EINVAL;
}
kv_params.nslots = params->nslots;
kv_params.datakiller = NULL;
switch (params->key_type) {
case KVMGR_KEY_INT:
kv_params.keyops = &int_key_ops;
break;
case KVMGR_KEY_STRING:
kv_params.keyops = &string_key_ops;
break;
default:
return -EINVAL;
}
kvm->kvtab = makeInstance(&kvtab_type, &kv_params);
if (kvm->kvtab == NULL) {
return -errno;
}
data_params.delta = params->delta;
data_params.datakiller = params->datakiller;
kvm->data = makeInstance(&xzarray_type, &data_params);
if (kvm->data == NULL) {
destroy(kvm->kvtab);
return -errno;
}
kvm->nslots = params->nslots;
kvm->key_type = params->key_type;
my_rwlock_init(&(priv->lock));
return 1;
}
static void xkvmgr_destroy(meta_obj_t* obj)
{
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
if (kvm->data != NULL) {
destroy(kvm->data);
}
if (kvm->kvtab != NULL) {
destroy(kvm->kvtab);
}
my_rwlock_fini(&(priv->lock));
memset(kvm, 0x0, sizeof(*kvm));
return;
}
static int xkvmgr_add(meta_obj_t* obj, void* key, void* value)
{
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
int rc;
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
int idx;
rc = vstore(kvm->data, value);
if (rc >= 0) {
idx = rc;
rc = key_add(kvm->kvtab, key, (void*)(long)idx);
if (rc >= 0) {
rc = idx;
} else {
vclear(kvm->data, idx);
}
}
return rc;
}
static int xkvmgr_find(meta_obj_t* obj, void* key, void** value)
{
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
int rc;
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
void* data;
*value = NULL;
rc = key_find(kvm->kvtab, key, &data);
if (rc >= 0) {
int idx = (int)(long)data;
rc = vget(kvm->data, idx, &data);
if (rc >= 0) {
*value = data;
rc = 1;
}
}
return rc;
}
static int xkvmgr_make_iter(iterator_t* iter, va_list ap)
{
meta_obj_t* obj = iter->obj;
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
iterator_t* it = calloc(1, sizeof(*it));
int rc = 0;
if (it == NULL) {
return -ENOMEM;
}
rc = iter_make(it, kvm->data);
if (rc < 0) {
FREEUP(it);
return rc;
}
iter->data = it;
return RTTI_ITER_OK;
}
static int xkvmgr_iter_next(iterator_t* iter, void** data)
{
iterator_t* it = (iterator_t*)iter->data;
return iter_next(it, data);
}
static int xkvmgr_iter_reset(iterator_t* iter)
{
iterator_t* it = (iterator_t*)iter->data;
return iter_reset(it);
}
static void xkvmgr_iter_destroy(iterator_t* iter)
{
iterator_t* it = iter->data;
meta_obj_t* obj = iter->obj;
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
if (it != NULL) {
iter_destroy(it);
FREEUP(it);
}
iter->data = NULL;
return;
}
static int xkvmgr_get_vec(meta_obj_t* obj, int count, int* idxs, void** data)
{
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
int rc = 0;
rc = vget_vec(kvm->data, count, idxs, data);
return rc;
}
static int xkvmgr_get(meta_obj_t* obj, int idx, void** data)
{
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
int rc = 0;
rc = vget(kvm->data, idx, data);
return rc;
}
static int xkvmgr_put_vec(meta_obj_t* obj, int count, int* idxs, void** data)
{
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
int rc = 0;
rc = vput_vec(kvm->data, count, idxs, data);
return rc;
}
static int xkvmgr_put(meta_obj_t* obj, int idx, void* data)
{
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
xkvmgr_private_t* priv = RTTI_OBJ_TO_PRIV(obj);
int rc = 0;
rc = vput(kvm->data, idx, data);
return rc;
}
static int xkvmgr_lock(meta_obj_t* obj, uint64_t option)
{
xkvmgr_private_t* priv = (xkvmgr_private_t*)RTTI_OBJ_TO_PRIV(obj);
int rc = -EINVAL;
switch (option) {
case RTTI_LOCK_READ:
rc = my_rwlock_rlock(&(priv->lock));
break;
case RTTI_LOCK_WRITE:
rc = my_rwlock_wlock(&(priv->lock));
break;
default:
break;
}
return rc;
}
static void xkvmgr_unlock(meta_obj_t* obj)
{
xkvmgr_private_t* priv = (xkvmgr_private_t*)RTTI_OBJ_TO_PRIV(obj);
my_rwlock_unlock(&(priv->lock));
return;
}
typeinfo_t xkvmgr_type = {
.name = "xkvmgr",
.privateSize = sizeof(xkvmgr_private_t),
.dataSize = (datasize_fn_t)xkvmgr_datasize,
.init = (obj_init_fn_t)xkvmgr_init,
.destroy = (obj_destroy_fn_t)xkvmgr_destroy,
.interfaces
= { .kv = &(intf_kv_t) { .add_key = (key_add_fn_t)xkvmgr_add,
.del_key = (key_del_fn_t)NULL,
.find_key = (key_find_fn_t)xkvmgr_find },
.varray = &(intf_varray_t) { .vstore = (vstore_fn_t)NULL,
.vget = (vget_fn_t)xkvmgr_get,
.vget_vec = (vget_vec_fn_t)xkvmgr_get_vec,
.vput = (vput_fn_t)xkvmgr_put,
.vput_vec = (vput_vec_fn_t)xkvmgr_put_vec,
.vclear = (vclear_fn_t)NULL,
.vfreeze = NULL, /* wired in P0-2 */
.vthaw = NULL /* wired in P0-2 */ },
.iter
= &(intf_iterator_t) { .make = (iter_make_fn_t)xkvmgr_make_iter,
.next = (iter_next_fn_t)xkvmgr_iter_next,
.reset = (iter_reset_fn_t)xkvmgr_iter_reset,
.destroy
= (iter_destroy_fn_t)xkvmgr_iter_destroy },
.lock = &(intf_lock_t) { .lock = (lock_fn_t)xkvmgr_lock,
.unlock = (unlock_fn_t)xkvmgr_unlock } }
};
/* ------------------------------------------------------------------ */
/* kvmgr_capture_keys — P0-14a */
/* ------------------------------------------------------------------ */
typedef struct {
slot_key_t* keys;
size_t capacity;
size_t filled;
kvmgr_key_type_t key_type;
int oom;
} capture_ctx_t;
static void capture_one(void* raw_key, int index, void* arg)
{
capture_ctx_t* ctx = (capture_ctx_t*)arg;
if (ctx->oom || ctx->filled >= ctx->capacity) {
return;
}
slot_key_t* sk = &ctx->keys[ctx->filled];
sk->index = index;
if (ctx->key_type == KVMGR_KEY_STRING) {
const char* src = (const char*)raw_key;
strncpy(sk->key, src, XCE_MAXNAME_LEN - 1);
sk->key[XCE_MAXNAME_LEN - 1] = '\0';
sk->key_len = strlen(sk->key) + 1;
} else {
/* Integer key — format as decimal string */
int64_t intval = (int64_t)(long)raw_key;
int n = snprintf(sk->key, XCE_MAXNAME_LEN,
"%" PRId64, intval);
if (n < 0 || n >= XCE_MAXNAME_LEN) {
/* Truncated — mark as oom so caller can detect */
ctx->oom = 1;
return;
}
sk->key_len = (size_t)n + 1;
}
ctx->filled++;
}
/*
* Count live entries by iterating; kvtab_foreach returns the count.
* We use a NULL callback to count first, then allocate, then fill.
*/
static void count_one(void* key, int index, void* arg)
{
(void)key;
(void)index;
size_t* n = (size_t*)arg;
(*n)++;
}
int kvmgr_capture_keys(var kvmgr_var,
slot_key_t** keys_out,
size_t* n_keys_out)
{
*keys_out = NULL;
*n_keys_out = 0;
if (kvmgr_var == NULL) {
return -EINVAL;
}
/* The public API pointer (var) points to the data region;
* convert to the meta_obj_t header, then extract xkvmgr_t. */
meta_obj_t* obj = RTTI_DATA_TO_OBJ(kvmgr_var);
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
/* Count live entries */
size_t n = 0;
kvtab_foreach(kvm->kvtab, count_one, &n);
if (n == 0) {
return 0;
}
slot_key_t* keys = calloc(n, sizeof(slot_key_t));
if (keys == NULL) {
return -ENOMEM;
}
capture_ctx_t ctx = {
.keys = keys,
.capacity = n,
.filled = 0,
.key_type = kvm->key_type,
.oom = 0,
};
kvtab_foreach(kvm->kvtab, capture_one, &ctx);
if (ctx.oom) {
free(keys);
return -ENOMEM;
}
*keys_out = keys;
*n_keys_out = ctx.filled;
return 0;
}
/* ------------------------------------------------------------------ */
/* kvmgr_reserve — job-array-single-wal-record task 7.1 (bug-1443) */
/* ------------------------------------------------------------------ */
int kvmgr_reserve(var kvmgr_var, int n)
{
if ((kvmgr_var == NULL) || (n < 0)) {
return -EINVAL;
}
/* Same public-pointer -> meta_obj_t -> xkvmgr_t unwrap as
* kvmgr_capture_keys() above. */
meta_obj_t* obj = RTTI_DATA_TO_OBJ(kvmgr_var);
xkvmgr_t* kvm = RTTI_OBJ_TO_DATA(xkvmgr_t, obj);
/* Pre-grow the value store ONLY -- this is the piece xkvmgr_add()
* mutates first (vstore(kvm->data, ...), see xkvmgr_add() above) and
* the only piece with an N-scaling realloc() reachable from a bulk
* key_add() loop. kvm->kvtab's own per-entry hash-chain-node allocation
* (kv.c) is deliberately NOT covered here -- see this function's doc
* comment (kvmgr.h) for why. */
return xzarray_reserve(kvm->data, n);
}